HAL Id: hal-02561845
https://hal.umontpellier.fr/hal-02561845
Submitted on 18 Nov 2020
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Highly siderophile and chalcophile element behaviour in
abyssal-type and supra-subduction zone mantle: New
insights from the New Caledonia ophiolite
Arianna Secchiari, Philipp Gleissner, Chunhui Li, Alexey Goncharov, Ralf
Milke, Harry Becker, Delphine Bosch, Alessandra Montanini
To cite this version:
Arianna Secchiari, Philipp Gleissner, Chunhui Li, Alexey Goncharov, Ralf Milke, et al.. Highly
siderophile and chalcophile element behaviour in abyssal-type and supra-subduction zone
man-tle: New insights from the New Caledonia ophiolite. Lithos, Elsevier, 2020, 354-355, pp.105338.
�10.1016/j.lithos.2019.105338�. �hal-02561845�
Research Article
Highly siderophile and chalcophile element behaviour in abyssal-type
and supra-subduction zone mantle: New insights from the New
Caledonia ophiolite
Arianna Secchiari
a,b, Philipp Gleissner
a, Chunhui Li
c, Alexey Goncharov
d, Ralf Milke
a, Harry Becker
a,
Delphine Bosch
e, Alessandra Montanini
b,⁎
aInstitut für Geologische Wissenschaften, Freie Universität Berlin, Germany
bDepartment of Chemistry, Life Sciences and Environmental Sustainability, University of Parma, Italy cDepartment of Geophysics and Planetary Sciences, University of Science and Technology of China, Hefei, China dDepartment of Geophysics, Saint Petersburg State University, Russia
eGéosciences Montpellier, Université de Montpellier, France
a b s t r a c t
a r t i c l e i n f o
Article history: Received 31 May 2019
Received in revised form 10 December 2019 Accepted 10 December 2019
Available online 16 December 2019
The New Caledonia Ophiolite hosts one of the largest obducted mantle sections worldwide, offering a unique op-portunity to investigate key mantle processes. The ophiolite comprises refractory harzburgites, locally overlain by mafic-ultramafic cumulates, and minor lherzolites. Previous geochemical studies indicated that the lherzolites are akin to abyssal-type peridotites, while the harzburgites underwent multiple melting episodes in MOR and supra-subduction zone environments, followed by late stage metasomatism.
In this work, Os isotopes, highly siderophile (HSE) and chalcophile element data are reported for the New Cale-donia peridotites, in order to constrain the behaviour of these elements in abyssal-type and fore-arc mantle. The variably serpentinised lherzolites (LOI = 6.4–10.7%) yield slightly subchondritic to suprachondritic initial Os isotopic compositions (187Os/188Os
i= 0.1273–0.1329) and subchondritic to chondritic Re/Os ratios (0.04–0.11).
The gently sloping HSE patterns with increasing depletion towards Au show concentrations in the range of other lherzolites from MOR or continental setting. Sulphur contents are high and variable (202–1268 ppm), and were likely increased during serpentinisation. By contrast, Se/Te ratios and concentrations are within the range of primitive mantle (PM) values, meaning that these elements were not significantly mobilised during serpentinisation.
Although displaying homogenous petrographic and geochemical features, the harzburgites are characterised by extremely heterogeneous Re\\Os and HSE compositions.
Type-A harzburgites exhibit subchondritic187Os/188Os
i(0.1203–0.1266) and low Re/Os ratios (0.01–0.04). The
strong IPGE-PPGE fractionations (PdN/IrN= 0.21–0.56), coupled with positive Pt anomalies and S-Se-Te
abun-dances often below the detection limit, suggest high melt extraction rates, resulting in sulphide consumption and Os\\Ru metal alloy stabilisation.
Type-B harzburgites possess strongly fractionated, Os-Ir-Pt poor (Os = 0.003–0.072 ng/g, Ir = 0.0015–0.079 ng/g) and Pd\\Re enriched patterns, associated with chondritic to suprachondritic measured187Os/188Os (0.127–0.153).
These characters are uncommon for highly depleted mantle residues. Interaction with an oxidised component does not appear as a viable mechanism to account for the IPGE-depleted patterns of type-B harzburgites, as calcu-lated oxygen fugacities are close to the FMQ buffer (Log ΔFMQ = 0.35 to 0.65).
The strikingly uniform mineralogical and geochemical features displayed by both harzburgite sub-types suggest that the different HSE patterns are not linked to their recent evolution, implying that subduction-related pro-cesses were superimposed on geochemical heterogeneous mantle domains, which exerted an important control on HSE behaviour during melt extraction and post melting metasomatism.
We propose that the HSE characters of the studied peridotites reflect the presence of a highly heterogeneous mantle source with a long term (N1 Ga) evolution, possibly linked to the Zealandia formation.
© 2019 Elsevier B.V. All rights reserved.
⁎ Corresponding author.
E-mail address:alessandra.montanini@unipr.it(A. Montanini).
https://doi.org/10.1016/j.lithos.2019.105338
0024-4937/© 2019 Elsevier B.V. All rights reserved.
Contents lists available atScienceDirect
Lithos
1. Introduction
Highly siderophile elements (HSE: PGE+ Au\\Re) are powerful geo-chemical tracers that can provide useful information for a variety of mantle processes, such as mantle melting, metasomatism and melt-fluid/mantle interaction (e.g.Ackerman et al., 2009;Lorand et al., 2008; Luguet et al., 2001, 2003, 2007). However, our knowledge concerning the behaviour of HSE in mantle source rocks of primitive arc magmas and the role of the subduction zone environment on HSE partitioning (i.e. hydrous melting, melt/fluid-mantle interaction) still remains quite fragmentary. Furthermore, although abundant HSE data are now available for different types of mantle peridotites, HSE data on fore-arc peridotites are remarkably scarce (e.g.Becker and Dale, 2016). The New Caledonia ophiolite (Peridotite Nappe) hosts one of the largest and best preserved mantle sections worldwide, providing an cellent opportunity to investigate upper mantle processes. The rock ex-posures are dominated by harzburgite tectonites bearing a supra-subduction zone affinity (Marchesi et al., 2009;Pirard et al., 2013; Secchiari et al., 2020;Ulrich et al., 2010). The main geochemical and iso-topic features of these rock-types reflect a complex polyphase evolu-tion, including several melting episodes in different geodynamic settings and subduction zone metasomatism (Marchesi et al., 2009; Secchiari et al., 2020;Ulrich et al., 2010). Minor spinel and plagioclase lherzolites, with compositions similar to abyssal peridotites, occur as discrete bodies in the north-western part of the island. The lherzolites record a different history compared to the extremely refractory harzburgites, as highlighted by their different geochemical signature (Secchiari et al., 2016;Ulrich et al., 2010).
In this work, a set of fully characterised peridotites (i.e. whole rock and in situ major and trace element contents, Sr-Nd-Pb isotopes) from New Caledonia (Secchiari et al., 2016, 2019, 2020) has been used to investigate Re\\Os, HSE and chalcophile element (S-Se-Te) systemat-ics. The main aims of this work are: 1) to examine the behaviour of these elements in the lherzolites (i.e. presumed abyssal peridotites) and in the ultra-depleted harzburgites, which may represent rocks from a former supra-subduction zone mantle wedge; 2) to constrain the behaviour of HSE and chalcophile elements during subduction zone processes.
2. Geological setting and petrological background
New Caledonia is a NW–SE elongated island located in the SW Pa-cific region, between the eastern margin of Australia and the Vanuatu archipelago (Figs. 1and2a). The island represents the emerged por-tion of the submarine Norfolk Ridge and it is composed by a mosaic of volcanic, sedimentary and metamorphic terranes, ranging in age from Permian to Miocene (Aitchison et al., 1995;Cluzel et al., 2001, 2012;Lagabrielle et al., 2013). These terranes were amalgamated dur-ing two major tectonic events: 1) an Early Cretaceous tectonic conver-gence phase (Paris, 1981) and 2) a Paleocene to Late Eocene subduction culminated in the obduction of the ophiolite. Both events were characterised by high-pressure low-temperature (HP-LT) meta-morphism in connection with plate convergence. New Caledonia can be sub-divided into four main geological domains (Cluzel et al., 2001; seeFig. 1): (i) the Basement units (pre-Late Cretaceous base-ment and Late Coniacian-to-Late Eocene sedibase-mentary cover), (ii) the
Fig. 1. a) Present-day structures of the Southwest Pacific region modified afterCluzel et al. (2012). Dark orange, land; light orange, continental plateau; white, oceanic basins (LHR: Lord Howe Rise, NR: Norfolk ridge, LR: Loyalty ridge, HP: Hikurangi Plateau); b) simplified geological map of New Caledonia showing distribution of the Peridotite massifs (modified afterCluzel et al., 2012). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 2. Variations of Os, Ru, Rh, Pt, Pd, Au and Re vs. Ir for the New Caledonia peridotites. Abyssal peridotites (Kane fracture zone:Snow and Schmidt, 1998;Brandon et al., 2000;Luguet et al., 2001, 2003;Marchesi et al., 2013; MAR:Harvey et al., 2006; Lena trough:Lassiter et al., 2014) and ophiolitic peridotites (IL-EL: Internal and External Ligurides,Snow et al., 2000;
Cenozoic HP-LT metamorphic belt, (iii) the basaltic Poya Terrane and (iv) a large slab of peridotites, named the Peridotite Nappe.
The Peridotite Nappe represents an allochtonous sheet of oceanic lithosphere belonging to the former South Loyalty basin thrust on the continental basement of the Norfolk Ridge at the end of the Eocene sub-duction. The emplacement of the ophiolitic nappe resulted from the failed subduction of the Norfolk Ridge tip in a NE-dipping subduction zone, which culminated in the obduction of the Loyalty subarc litho-sphere ~ 34 Ma ago (Cluzel et al., 2012).
The Peridotite Nappe has an extension of about 8000 km2and is
mostly exposed in the Massif du Sud, where a thick harzburgite– dunite unit, locally overlain by kilometre-scale lenses of mafic and ul-tramafic intrusives, crops out. The sequence is believed to represent a crust-mantle boundary that records the onset of the Eocene subduc-tion in a nascent arc setting (Marchesi et al., 2009;Pirard et al., 2013;Secchiari et al., 2018). Recent geochemical studies have shown that the ultramafic intrusives (i.e. dunites and wehrlites) crystallised from variably depleted melts with island arc basalt affinity, after mas-sive interactions with the underlying harzburgite (Marchesi et al., 2009; Pirard et al., 2013). In contrast, the mafic rocks (i.e. gabbronorites) have a cumulate origin (Marchesi et al., 2009;Pirard et al., 2013;Secchiari et al., 2018) and derive from crystallisation of primitive, non-aggregated, ultra-depleted melts showing involvement of a subduction-related component in their source (Secchiari et al., 2018).
The harzburgites are also exposed in the northern Tiébaghi massif (Ulrich et al., 2010) or as sparse tectonic klippen in the central part of the island (e.g. Kopeto, Poya, Koniambo), where exceptionally fresh pe-ridotites display primary mineral assemblages similar to the more serpentinised rocks of the Massif du Sud.
The New Caledonia harzburgites bear an overall ultra-depleted com-position, inherited from a complex multistage evolution linked to the development of the Eocene subduction system (Marchesi et al., 2009; Secchiari et al., 2020;Ulrich et al., 2010). Geochemical studies have pro-posed that the harzburgites formed by high degrees of fluid-assisted melting (up to 20–25% in a supra-subduction zone environment, see Marchesi et al., 2009;Ulrich et al., 2010). More recently, the work of Secchiari et al. (2020)provided further constraints on the evolution of these rock-types, tracking their history from melting to late stage meta-somatism. Accordingly, the harzburgites underwent two partial melting episodes in the spinel stability field: a first melting phase in a MOR
setting (15% melting degrees), followed by hydrous melting in a supra-subduction zone setting (up to 18% fluid-assisted melting). Post-melting cooling and re-equilibration at lithospheric conditions was accompanied by interaction with ultra-depleted slab-derived hy-drous melts (Secchiari et al., 2019, 2020). These metasomatic processes are testified in the harzburgites by the widespread occurrence of sec-ondary metasomatic phases (i.e. thin films of Al2O3-, CaO- poor
orthopyroxene, and low Al2O3and Na2O clinopyroxene), L-MREE and
Zr\\Hf bulk rock enrichments, and possibly by the poorly radiogenic Nd isotopic ratios shown by some samples (Secchiari et al., 2020).
Compared to the harzburgites from the central and the southern massifs, Tiébaghi samples display a more fertile nature, as revealed by the higher bulk trace element concentrations as well as by the occur-rence of a small fraction (up to 4 vol%) of clinopyroxene (seeUlrich et al., 2010;Secchiari, 2016PhD thesis).
The main geochemical and petrological features of the spinel and plagioclase lherzolites are thought to reflect moderate melt extraction degrees (8–9%) in a MOR environment, followed by refertilisation by depleted MORB-type melts, yielding plagioclase lherzolites. The main petrological and geochemical features of the lherzolites have been re-ported in detail bySecchiari et al. (2016).
2.1. Sample description
In this contribution, sixteen samples of peridotites fully characterised for lithophile element geochemistry (i.e. major, trace ele-ment and Sr-Nd-Pb isotope compositions) were analysed for mass frac-tions of all PGE, Re, Au, S, Se, Te and187Os/188Os. Detailed descriptions of
the lherzolites and harzburgites, including trace element chemistry and Sr-Nd-Pb isotopes, are provided in the works ofSecchiari et al. (2016) andSecchiari et al. (2019, 2020), respectively.
Lherzolite samples are from the Poum and Babouillat areas, while the harzburgites were collected from several outcrops and mine zones along the island: Yaté, Kopeto, Poya, Poro and Tiébaghi (Fig. 1b and Table 1). The lherzolites include slightly serpentinised (LOI = 6.9–10.7%) spinel and plagioclase lherzolites, while the harzburgites are typically not or only weakly serpentinised (LOI = 0–3%), except for samples YA1, TI1 and TI2 (LOI = 6.0–9.0%).
Both lherzolites and harzburgites are low strain mantle tectonites, showing dominant porphyroclastic textures (Fig. S1a-b) and local protomylonite development. Spinel lherzolites have 7–8 vol%
Table 1
Concentrations of the HSE, S, Se and Te, Os isotopes, selected major elements and ratios for the New Caledonia peridotites.
Sample Rock Al2O3 LOI Os Ir Ru Rh Pt Pd Au Re OsN/IrN OsN/RuN PdN/IrN RuN/IrN
wt% % (ng/g) (ng/g) (ng/g) (ng/g) (ng/g) (ng/g) (ng/g) (ng/g) POU1A Pl L 2.49 7.56 2.93 2.48 5.31 1.57 5.54 6.23 0.704 0.319 1.06 0.99 1.24 1.07 POU2 Sp L 1.46 8.63 3.88 3.45 6.76 1.23 6.60 6.56 1.04 0.213 1.01 0.18 0.94 5.71 POU2B Sp L 2.57 10.69 3.31 2.88 5.93 1.04 6.49 5.49 0.503 0.165 1.03 1.09 0.94 0.95 POU3 Sp L 1.30 9.60 2.91 2.58 5.07 0.86 5.15 4.14 0.642 0.234 1.01 1.05 0.79 0.97 BA1 Sp L 1.74 6.39 3.34 2.85 5.95 1.05 5.62 5.51 0.943 0.129 1.06 0.27 0.96 3.88 BAB1B Pl L 2.61 6.98 2.96 2.61 5.46 0.965 6.73 5.42 0.641 0.125 1.02 0.24 1.02 4.30 BAB2B Pl L 2.83 8.43 3.00 2.57 5.08 1.16 6.06 5.33 1.33 0.138 1.05 1.67 1.02 0.63 TI1 H 0.78 9.01 1.51 0.864 2.52 0.275 2.42 0.488 1.25 0.022 1.57 1.08 0.28 1.46 TI2 H 1.03 6.04 2.07 1.14 3.78 0.254 2.70 0.499 0.264 0.019 1.63 0.98 0.21 1.65 PO4 H 0.43 0.18 1.07 0.499 1.53 0.191 2.56 0.570 0.160 0.018 1.93 1.26 0.56 1.53 YA1 H 0.46 6.83 0.554 0.297 2.35 0.326 0.268 0.263 1.17 0.022 1.67 0.42 0.44 3.95 PY1 H 0.78 0.00 0.049 0.077 0.643 0.269 2.08 0.378 0.336 0.016 0.58 0.14 2.43 4.19 Duplicate 0.023 0.133 0.693 0.171 2.44 0.588 0.311 0.062 KPT2 H 0.70 3.03 0.004 0.022 0.323 0.057 0.078 0.431 0.107 0.018 0.18 0.02 9.50 7.22 Duplicate 0.036 0.017 0.211 0.040 0.210 0.620 0.144 0.010 KPT5 H 0.74 0.12 0.004 0.066 0.165 0.048 0.038 0.464 0.654 0.024 0.05 0.04 3.46 1.25 Duplicate 0.007 0.062 0.165 0.054 0.046 0.512 0.603 0.056 PO3 H 0.41 0.00 0.003 0.015 0.256 0.023 0.006 0.428 0.084 0.012 0.18 0.02 14.12 8.57 Duplicate 0.055 0.018 0.243 0.016 0.009 0.573 0.293 0.008 KPT3 H 0.67 0.67 0.072 0.079 0.280 0.172 0.248 0.250 0.288 0.410 0.81 1.87 0.17 0.19
clinopyroxene and display a typical abyssal-type REE signature. The pla-gioclase lherzolites show melt impregnation microstructures (Fig. S1b) and are slightly enriched in incompatible trace element (REE, Ti, Y, and Zr) with respect to the spinel lherzolites. Harzburgites are extremely de-pleted rocks, as highlighted by the general absence of clinopyroxene (with the exception of sample TI2, where clinopyroxene is ~ 4 vol%, Fig. S1c) and remarkably low incompatible trace element contents (Secchiari et al., 2020). The primary mantle assemblage is composed of olivine, orthopyroxene and spinel. The occurrence of thin films of metasomatic ortho- and clinopyroxene (Fig. S1d) was interpreted as resulting from percolation of small fractions of subduction-related magmas (Secchiari et al., 2019, 2020).
All the lherzolites studied here contain trace amounts of sulphides with variable size, shape and position. Sulphides in five selected sam-ples (i.e. BA1, POU2, POU3, POU1A, BAB1B, see Table S1) were analysed for their major element chemistry composition.
Frequent interstitial sulphide grains occur as polyhedral blebs, rounded or irregular-shaped grains (Fig. S2a-d-e), generally located at olivine-orthopyroxene or at olivine-olivine grain boundaries. They range from tiny crystals (~ 60 × 30 μm) up to 300 × 100/150 μm in max-imum dimensions. Enclosed sulphides hosted within large olivine porphyroclasts have been also observed. Sulphide inclusions (Fig. S2a-e) have variable size (30 to 80 μm × 60 to 150 μm) and shape, from poly-hedraltospherical.Very tinyenclosed sulphide grains(~5–6 μm×10μm) have been recognised in sample POU1A.
Major element composition of the sulphide phases is relatively ho-mogeneous (Table S1 and Fig. S2, S3), showing no significant difference among enclosed and interstitial sulphides. Chemical composition of the studied sulphides mainly fall in the field of Ni-poor (Fe/Ni = 1.3–2.1) monosulphide solid solution + Liquid (Fig. S3), but Ni-rich (Fe/Ni = 0.7–0.8) pentlandite, resulting from cooling and re-equilibration (Guo et al., 1999), has been also identified in four of the investigated samples (POU2, POU3, POU1A, BAB1B).
In the harzburgites, sulphide phases have not been identified, nor through petrographic observation or microprobe analyses. This is con-sistent with the refractory nature and the high melt extraction rates ex-perienced by these rock-types, which prevented sulphide retention in the residual mantle source.
3. Analytical methods
3.1. HSE and chalcophile elements
Seven lherzolites and nine harzburgites (including four duplicates) have been analysed in the geochemistry laboratory at Freie Universität for HSE, S, Se, Te mass fractions in whole rocks and187Os/188Os.
Detailed procedure descriptions have been given in previous works from this laboratory (e.g.Fischer-Gödde et al., 2011;Wang et al., 2013; Wang and Becker, 2013). The methods will only be briefly summarised below.
About 2.5 g of sample powder was weighed into 90 mL quartz glass digestion vessels and spiked with mixed 191Ir–99Ru–194Pt–105Pd, 77Se\\125Te,185Re\\190Os and34S solutions. Then, 5 mL 14 mol/L, N
2
-bubbled HNO3and 2.5 mL 9 mol/L HCl were added. The vessels were
im-mediately sealed with Teflon tape and samples were digested for 16 h at 320 °C and 100 bar. After digestion, osmium was extracted from the re-verse aqua regia into chloroform, back extracted into HBr (Cohen and Waters, 1996), and further purified by micro distillation from a H2SO4-dichromate solution into 15 μL of HBr (Birck et al., 1997)
Osmium isotopes were determined as OsO3−in negative mode using
the Thermo Finnigan Triton TIMS, using a secondary electron multiplier. Signal intensities of the spike isotope 190Os of samples were
~150,000–500,000 cps. Standard runs with different amounts of Os on the filament (10 pg and 100 pg) were also run in between the studied samples, yielding an average value of 0.1139 ± 0.0002 (2 s. d., n = 24) for 100 pg loads. Two hundreds scans were collected in each mea-surement for high-Os samples, while at least 120–140 scans were ob-tained for the low-Os samples. Raw data were corrected for isobaric OsO3−interferences, mass fractionation using the192Os/188Os ratio of
3.08271, contributions from the Os spike solution and blank contribu-tions.187Os/188Os were finally adjusted relative to the mean of the Os
standard. The oxygen isotope compositions used for the oxide correc-tion of Os oxide molecules were18O/16O of 0.00204 and17O/16O of
0.00037 (Nier, 1950).
About 50% of the digestion solution was used for separation of the HSE fraction and about 30% for S–Se–Te separation. Chemical separation of the HSE fraction from the matrix was performed on columns filled
PtN/IrN PtN/RuN 187Re/188Os
(2SE)
187Os/188Os 2SE 187Os/188Os i γOs (53Ma) TMA (PM) T(PM)RD T(PM)RD2 S Se Te S/Se Se/Te measured Ga Ga Ga (μg/g) (ng/g) (ng/g) 1.03 0.96 0.525(1) 0.130822 9.1E-06 0.130358 2.9 0.8 f f 202 74.9 11.5 2703 6.5 0.88 0.15 0.265(1) 0.127574 6.8E-06 0.127340 0.5 0.7 f 0.3 582 69.6 11.6 8364 6.0 1.04 1.09 0.240(1) 0.128227 8.0E-06 0.128016 1.1 0.4 f 0.2 1268 77.8 10.8 16,289 7.2 0.92 0.95 0.387(1) 0.131356 9.1E-06 0.131015 3.4 f f f 362 54.0 7.3 6691 7.4 0.92 0.24 0.186(1) 0.131485 7.9E-06 0.131320 3.7 f f f 294 91.3 13.8 3218 6.6 1.19 0.28 0.202(1) 0.129664 8.9E-06 0.129485 2.2 f f 0.0 327 66.2 9.5 4943 7.0 1.09 1.73 0.223(1) 0.133084 8.9E-06 0.132887 4.9 f f f 289 67.6 9.8 4278 6.9 1.29 0.89 0.070(25) 0.12479 1.7E-05 0.12473 −1.5 0.8 0.7 0.7 9 1.1 1.0 8188 1.0 1.09 0.66 0.045(18) 0.12662 1.3E-05 0.12658 −0.1 0.5 0.4 0.4 10 0.7 0.4 14,053 1.7 2.36 1.54 0.081(35) 0.12040 2.4E-05 0.12033 −5.0 1.5 1.3 1.3 6 3.1 0.8 2084 4.1 0.42 0.11 0.196(68) 0.12551 4.7E-05 0.12534 −1.0 1.0 0.6 0.6 40 4.9 0.4 8117 11.4 2.97 2.97 1.62(8) 0.1299 5.3E-04 0.1284 1.4 0.0 0.0 0.2 6 0.9 0.9 6103 1.1 13.0(1) 0.127 1.0E-03 0.115 −8.9 53 0.3 1.0 196,197 0.3 0.22 0.22 19(1) 0.148 5.6E-03 0.131 3.6 0.1 f f 5 1.3 0.4 4164 2.8 1.4(1) 0.302 1.3E-03 0.301 137.8 7 1.3 8.2 4947 0.2 0.21 0.21 32(2) 0.147 6.2E-03 0.118 −6.6 0.0 f f 3 3.0 0.7 1133 4.1 37(1) 0.160 3.2E-03 0.127 0.5 3 0.7 1.3 4453 0.5 0.02 0.02 19(2) 0.153 8.4E-03 0.136 8.0 0.1 f f 3 1.4 0.6 2129 2.2 0.69(7) 0.1239 4.7E-04 0.1233 bdl bdl 0.8 – – 6.92 6.92 28(1) 0.1273 3.5E-04 0.1204 −5.1 0.0 0.3 1.3 6 bdl bdl – – Duplicate: replicate digestion of the same sample powder.
Pl L = plagioclase lherzolite, Sp L = spinel lherzolite, H = harzburgite.
Values of PM187Os/188Os = 0.1296 and187Re/188Os = 0.434 used for calculation of T
MAand TRDages (Meisel et al., 2001); f = future model ages. TRD2(PM) indicates depletion ages
with 10 mL of pre-cleaned Eichrom 50 W-X8 (100–200 mesh) cation exchange resin (Fischer-Gödde et al., 2011). During separation, the HSE fraction was collected in 14 mL 0.5 mol/L HCl-40 vol% acetone mix-ture. After the volume of the solution has been reduced to about 2 mL it was analysed for Au, Re, Ir and Pt. In order to remove interfering Cd, the remaining solution was further purified in 0.2 mol/L HCl on 3 mL Eichrom 50 W-X8 (100–200 mesh) resin. The collected solution was evaporated to near dryness and the residue was taken up in 0.28 M HNO3for ICP-MS analysis. The analyses were carried out using a single
collector Element XR instrument. We used either a Scott-type spray chamber (Re, Ir, Pt, Au) or an Aridus-I desolvator (Ir, Ru, Pt, Rh, Pd) at an oxide formation rate of CeO+/Ce+b0.004.
A two-step ion exchange chromatography method was used for sep-aration of S, Se and Te (seeWang et al., 2013). Sulphur measurements were performed on the S\\Se fraction at medium mass resolution mode on the Element XR. Selenium and Tellurium were measured using a double pass Scott type glass spray chamber at low mass resolu-tion mode on the Element XR, combined with a hydride generaresolu-tion sample introduction system by reacting the sample solution with 1 g/ 100 g NaBH4in 0.05 mol/L NaOH (seeWang et al., 2013for details).
For each batch of analysis, one procedural blank has been used. Pro-cedural blanks yielded the following mean values (± 1 s.d., n = 4–5): Re = 2.5 ± 2.0 pg; Os = 0.5 ± 0.3 pg with187Os/188Os ratios of
0.14 ± 0.03; Ir = 15 ± 5 pg; Ru = 45 ± 14 pg; Rh = 24 ± 22 pg; Pt = 23 ± 29 pg; Pd = 640 ± 330 pg; Au = 4 ± 2 pg; Te = 1.1 ± 0.8 ng; Se = 2.3 ± 0.8 ng; S = 2.8 ± 0.7 μg. Samples were corrected for total procedural blanks using the mean values. Blank corrections for Re are negligible for most of the analysed samples (≤0.3–0.8%), but more significant for the harzburgites (~4–8%). Blank corrections for Pt and Pd are again negligible for the lherzolites (~ 0.2–0.3%), a few per-cents for the harzburgites (~0.4–4%, with the exception of KPT2, KPT5 and PO3 for which the correction for Pt is ~ 11–36%). Blanks of Os, Ir, Ru and Rh are insignificant for most of the samples (≤0.4%) but higher for the most depleted harzburgites, i.e. KPT2, KPT5 (~2–7% for Os, Ir, Rh) and PO3 (~ 9% for Os and Ir, 13% for Rh). Blank corrections for S, Se and Te in lherzolites range between 1 and 1.7% (S - Se) and 3–6% (Te), while for the harzburgites blank corrections for these elements strongly affected the obtained results (corrections ~10–26% for S and up to 40–80% for Se and Te), given the very low measured abundances.
3.2. Oxygen fugacity
Iron oxidation state in studied spinels was measured using the “flank method” developed for the JEOL JXA-8200 electron microprobe at Freie Universität (Goncharov, 2018). The position shift and intensity variation of FeLα and FeLβ x-ray lines were investigated using as standards a col-lection of nineteen natural mantle spinels previously characterised by Mössbauer spectroscopy for their Fe3+/ΣFe at IPGG RAS (St. Petersburg,
seeGoncharov and Ionov, 2012;Goncharov et al., 2015).
The analytical procedure was similar to the experiments performed over the last decades to study Fe3+/ΣFe in mantle garnets after the
“flank” approach developed byHöfer and Brey (2007).
The flank positions for the spectrometer were obtained from the dif-ference spectrum of almandine and andradite with known iron oxida-tion state in the wavelength range related to FeLα and FeLβ lines. The intensities at flank positions near FeLα and FeLβ lines were collected as two fake elements, setting counting time at 300 s on both measure-ments, with 3 repetitions in the core and in the rim respectively, and considering 5 separate spinel grains within one thin section. Measure-ment conditions were 15 kV and 60 nA using TAP crystal for intensities at flank positions and with the remaining four spectrometers measuring chemical composition at the same spot simultaneously. The studied spi-nels show no significant core to rime zoning in terms of FeO content, Cr# and FeLβ/FeLα ratio. Averaged FeLβ/FeLα ratios for each sample were used to calculate iron oxidation state of spinel using the equation
with coefficients obtained after investigation of the standard collection (seeTable 2).
4. Results
4.1. HSE and chalcophile elements in spinel and plagioclase lherzolites HSE and chalcophile element compositions of the New Caledonia lherzolites are reported inTable 1and displayed inFigs. 2, 3, 4. Spinel and plagioclase lherzolites are relatively homogeneous in terms of HSE, Se, Te concentrations, abundance patterns and Os isotopic compo-sitions, with plagioclase-bearing samples showing indistinguishable patterns from those of spinel lherzolites. The concentrations of the highly siderophile and chalcophile elements are in the range of those observed for modern abyssal and ophiolitic peridotites, displaying good correlation for Ir group PGE (IPGE, e.g., Os vs. Ir and Ir vs. Ru) and more dispersed variations for the Pt group PGE (PPGE, Fig. S4). In primitive mantle (PM) normalised concentration diagrams (Fig. 3), the lherzolites exhibit flat or gently sloping negative patterns with sim-ilar PM-normalised PGE concentrations and depletions in Au (except for sample BAB2B) and Re compared to the PGE (with the exception of POU1A and POU3 for Re). Overall, absolute contents of the PGE are sim-ilar or slightly lower than primitive mantle (PM) values (Becker et al., 2006;Fischer-Gödde et al., 2011), overlapping the field of the abyssal peridotites and peridotite tectonites from continental settings (e.g.,Fig. 2andBecker and Dale, 2016). Ru/Ir and Pd/Ir ratios are suprachondritic, as observed for other mantle lherzolites (e.g.Becker et al., 2006;Lorand et al., 1999;Luguet et al., 2003;Rehkämper et al., 1999).
Initial187Os/188Os ratios calculated at 53 Ma (i.e. the inferred age of
initial magmatism in the subduction system, e.g.Cluzel et al., 2006) vary from chondritic to slightly suprachondritic (0.1273–0.1329,Fig. 4a), corresponding to γOs(53Ma)of 0.5 to 4.9. These values overlap with
data of abyssal peridotites and orogenic peridotites, but tend to be somewhat higher than for other mantle lherzolites bearing comparable depletion degrees (Fig. 4).187Re/188Os ranges from subchondritic to
slightly suprachondritic values (0.186–0.525, seeFig. 4c).
Se and Te are positively correlated in the lherzolites (Fig. 5a) and range between 54 and 91.3 ng/g and 7.3–13.8 ng/g, r espectively. Se/ Te ratios (5.9–7.1) are slightly lower than the PM value and similar to the data previously obtained on depleted lherzolites (Wang and Becker, 2013). Se and Te do not display any correlation with PGE abun-dances, with the exception of Te, which shows a weak correlation with Pd (Fig. 5b). S contents are high and variable (202–1268 μg/g) com-pared to unserpentinised peridotites, leading to high S/Se ratios (2703–16,289,Fig. 5c).
4.2. HSE and chalcophile elements in harzburgites
On the basis of HSE behaviour and Os isotopic compositions (Figs. 2, 3 and 6,Table 1), the studied harzburgites can be grouped into two dif-ferent sub-types: type-A and type-B.
Type-A harzburgites(TI1, TI2, PO4 and YA1) are characterised by no-tably lower PGE and chalcophile elements mass fractions (Figs. 2, 4) and more fractionated patterns (Fig. 6a) compared to the lherzolites. Mass fractions of the PGE are 1.07–2.07 ng/g for Os, 0.50–1.14 ng/g for Ir, 1.53–2.52 ng/g for Ru, 0.19–0.27 ng/g for Rh, 2.42–2.70 ng/g for Pt and 0.49–0.57 ng/g for Pd. Among the sub-group A, sample YA1 displays dis-tinct PGE abundances, showing much lower Os, Ir, Pt, Pd contents (0.55 ppb for Os, 0.30 ppb for Ir, 0.27 ppb for Pt and 0.26 ppb for Pd), with exceptions for Ru, Rh, Au and S.
HSE and chalcophile element diagrams of type-A harzburgites dis-play fractionated patterns, with concentrations decreasing from Os to-wards Re. Os and Ru are enriched compared to Ir, leading to correlated suprachondritic Os/Ir and Ru/Ir ratios (Os/Ir = 2.9–7.9; Ru/Ir = 1.8–2.1, Fig. 4). Pt and Au generally show positive spikes, more
pronounced for Au, with the exception of YA1, which displays a nega-tive Pt anomaly. Pd contents are low (b 0.1 PM values) and constant for PO4, TI1, TI2, with Pd/Ir showing subchondritic ratios for all the stud-ied samples. Positive correlations are observed between IPGE (Fig. 2 a-b-c) and Pt\\Ir (not shown), and, somewhat surprisingly, between IPGE and the fertility indicators (i.e. Al2O3and CaO, not shown). Mass
frac-tions of Te, Se and S are low, often close to or below the detection limit, again with the exception of the harzburgite YA1.
For all type-A harzburgites, Re concentrations are very low (about 0.02 ng/g), leading to subchondritic187Re/188Os ratios (0.045 to 0.196,
the latter value also reflecting the low Os concentrations in YA1). Os iso-topic compositions are subchondritic to chondritic (0.1203–0.1266, cor-responding −5 ≤ γOsi(53 Ma)≤ −0.1) and do not define any correlation
with187Re/188Os or incompatible element depletion indices (i.e. Al 2O3,
seeFig. 4).
Type-B harzburgitescomprise very fresh samples from Kopeto (KPT2, KPT3, KPT5), Poro (PO3) and Poya (PY1) massifs. Compared
to type-A harzburgites, these samples have much lower HSE abun-dances and display variable and strong fractionationsFig. 3among PGE and more incompatible chalcophile elements (Fig. 2 and 6b). In detail, Os, Ir and Pt show positive correlations (Fig. 2) and are strongly depleted compared to Ru, Rh and Pd (Os = 0.003–0.071 ppb, Ir = 0.015–0.079 ppb, with Os/Ru = 0.01–0.26 and Ru/Ir = 2.5–20). For sample PY1, Pt is enriched relative to IPGE, Rh and Pd (Pt/Rh = 7.7; Pt/Pd = 5.5). Pd, Re and S-Se-Te have similarly low normalised abundances, with chalcophile ele-ments often close to or below the detection limit. Au exhibits positive spikes for PY1, KPT3 and PO3 samples and tends to be more enriched than similar incompatible chalcophile elements (i.e. Pd and Re).
Measured187Os/188Os ratios vary from chondritic to suprachondritic
(0.1273–0.1534) and are coupled with high and variable187Re/188Os
(1.62–32). Initial Os isotopic compositions calculated at 53 Ma range from depleted to slightly suprachondritic values (0.1181–0.1365, −7 ≤ γOs(53Ma)≤ 3).
Table 2
Equilibration temperature, pressure, and oxygen fugacity calculated for selected peridotite samples. Average spinel compositions are also reported.
Sample Type Fe3+/ΣFea Fe3+/Σfe (rel. err.%)b T Ol-Spl (°C)c P (GPa) Woodd(1990)
log f(O2) ΔFMQ TI2 Type-A harz 0.21 (0.03) 15.0 870 1.0 −11.37 0.92 YA1 Type-A harz 0.18 (0.03) 14.5 815 1.0 −12.66 0.67 KPT5 Type-B harz 0.21 (0.03) 16.0 840 1.0 −12.20 0.65 PO3A Type-B harz 0.20 (0.03) 14.8 940 1.0 −10.76 0.33 PY1B Type-B harz 0.21 (0.03) 16.6 930 1.0 −10.90 0.35
Spinel
Sample Type SiO2 TiO2 Al2O3 Cr2O3 FeO Fe2O3 MnO MgO Total Cr# Mg# Lβ/Lαe
TI2 Type-A harz 0.04 0.03 43.45 25.35 10.20 2.99 0.15 16.93 99.13 0.281 0.747 0.804 YA1 Type-A harz 0.02 0.01 21.56 45.88 16.76 4.03 0.29 10.81 99.35 0.588 0.535 0.896 KPT5 Type-B harz 0.01 0.01 29.66 41.41 11.31 3.31 0.21 14.41 100.33 0.484 0.695 0.825 PO3A Type-B harz 0.02 0.00 17.27 53.48 12.58 3.54 0.25 12.32 99.47 0.675 0.636 0.833 PY1B Type-B harz 0.01 0.01 29.35 41.87 11.20 3.32 0.20 14.42 100.38 0.492 0.696 0.826 Fe3+/(Fe2++Fe3+) ratio has been calculated using the equation: Lα/Lβ x 0.42+ Cr# x 0.03 - (FeO + (Fe
2O3/1.113)) x 0.01 obtained after standardisation. aFe3+/(Fe2++Fe3+) in spinel determined by “flank”-method.
b Relative error in % calculated as a total of measurement errors for Fe, Cr, Al and X-ray intensities at flank positions. c T Ol-Spl = Equilibration temperature calculated using Li et al. (1995) formulation.
d Oxygen fugacity estimates are from Wood et al. (1990) and are reported as Δlog fO2 from the quartz-fayalite-magnetite (FMQ) buffer using the Fe3+/ΣFe values of the analysed
spinels. See text for further details.
e Current normalised ratio of the intensities at flank position for Lα and Lβ Fe lines.
Fig. 3. Primitive mantle normalised HSE and chalcogen patterns for the New Caledonia spinel and plagioclase lherzolites. Grey shaded area includes oceanic lherzolites from Mid-Atlantic and South West Indian ridges (Snow and Schmidt, 1998;Luguet et al., 2001;Luguet et al., 2003) and ophiolitic lherzolites from the Ligurian Units (Fischer-Gödde et al., 2011;Luguet et al., 2004;Snow et al., 2000). Normalising values afterBecker et al. (2006),Fischer-Gödde et al. (2011)andWang and Becker (2013).
Replicate analyses of samples PY1, KPT2, KPT5, PO3 yield quite sim-ilar results for Ru (b 5% relative deviation, except for sample KPT2) and Au (~ 6% for PY1 and KPT5) and acceptable results for Pt for PY1-KPT5 (11.5–14.0% relative deviation). Values appear much more scattered for Os and Re (RSD N 30%) and less dispersed for Rh and Pd (7 ≤ RSD% ≤32). The relative deviation of chalcophile elements is more limited, mostly b15%.
The large variations of mass fractions of HSE and chalcophile ele-ments in duplicate samples reflects the very low mass fractions of these elements combined with the inhomogeneity in the distribution of HSE carrier phases in gram-size quantities of rock powder, an issue that has already been recognised in peridotitic rocks (e.g.Becker et al., 2006;Luguet et al., 2007).
4.3. Oxybarometry
Results for calculations of fO2using coexisting olivine, orthopyroxene
and spinel (e.g., Wood, 1990, seeTable 2and Table S2) with the reaction are given inTable 2.
6 Fe2SiO4þ O2➔3 Fe2Si2O6þ 2 Fe3O4 ð1Þ
The calculation of fO2depends on the pressure and temperature of
equilibrium. For this study, we used olivine-spinel geothermometry (Li et al., 1995) as this thermometer is based on the same elements and minerals (Fe and Mg content in olivine and spinel) that are used to calculate fO2. As spinel peridotites lack a good barometer, in order
Fig. 4. a) Al2O3(wt%)-187Os/188Osiand b) Os-187Os/188Osiand c)187Re/188Os-187Os/188Os diagrams showing data from New Caledonia lherzolites and type-A harzburgites in comparison to
PM compositions. PM data fromMeisel et al., 1996. Data for abyssal peridotites are from Harvey et al., (2006) for Atlantic peridotites, Lassiter et al., (2014) for Lena through andLiu et al. (2015)for Gakkel ridge. SeeFig. 2for ophiolitic peridotites references.
to maximise consistency with temperatures determined with Li et al. (1995) thermometer, we have assumed a pressure of 1.0 GPa.
Ferric to total iron ratios (Fe3+/ΣFe) in spinel are similar within the
harzburgites, covering a limited range of values (0.18 ≤ Fe3+/Fe3++
Fe2+≤ 0.21, seeTable 2) indicating similar f
O2values for the two
harzburgite sub-groups (ΔlogFMQ = 0.35–0.92).
InFig. 7a-b spinel Cr# is reported against Fe3+/ΣFe ratio and
ΔlogFMQ for the analysed samples. Literature data for mantle perido-tites from different geodynamic settings (i.e. MOR and supra-subduction zone) and arc xenoliths have also been plotted for compar-ison. Overall, the harzburgites share remarkable similarities with supra-subduction zone peridotites from the Izu-Bonin area (Parkinson and
Fig. 5. a) Te vs. Se, b) Te vs. Pd and c) S vs. S/Se correlation diagrams for New Caledonia spinel and plagioclase lherzolites. Data for orogenic lherzolites fromWang and Becker (2013)are also plotted for comparison.
Pearce, 1998). Slightly lower Cr# values compared to the other harzburgites are displayed by TI2 sample, which in turn falls between abyssal peridotites and arc xenoliths array (Fig. 7a).
5. Discussion
5.1. HSE and Re\\Os systematics of the lherzolites
Major element composition and lithophile trace element chemistry of spinel lherzolites indicate moderately depleted compositions, inherited from moderate partial melting degrees (8–9%) of a DMM source, whereas plagioclase lherzolites originated through reactive melt percolation of spinel lherzolites by highly depleted, incremental melt fractions of a DMM source in the shallow lithosphere (Secchiari et al., 2016). In the following sections, the processes that may have af-fected HSE and Os isotopic signature of the New Caledonia lherzolites will be discussed: low temperature alteration, in particular serpentinisation, partial melting and the role of melt infiltration and chemical disequilibrium of the HSE in mantle rocks.
5.1.1. Effects of serpentinisation on HSE and187Os/188Os
Serpentinisation is a widespread process of hydrothermal alteration in ultramafic lithologies. However, its influence on HSE behaviour has not been investigated with much detail, despite some authors have pro-posed it as a possible cause for187Os ingrowth and Re variations in the
upper mantle (Snow and Reisberg, 1995;Standish et al., 2002;Walker et al., 1996). Recent experimental studies have shown that during serpentinisation the formation of secondary sulphides, Fe\\Ni alloys and native metals (Au\\Cu) is promoted by reducing fO2conditions
(Klein and Bach, 2009;Foustoukos et al., 2015) and thus, with the ex-ception of Au, HSE may be retained in the host rock. Comparison of par-tially serpentinised and unserpentinised peridotites displaying similar major element features supports the notion that at least PGE ratios are not significantly affected by moderate to strong serpentinisation (Becker and Dale, 2016).
The New Caledonia lherzolites underwent intermediate serpentinisation degrees (LOI = 6.4 to 10.7%, see paragraph 2.1 and Table 1), which had limited effects on the budget of fluid immobile moderately incompatible lithophile trace elements in these rocks
Fig. 6. a) Primitive mantle normalised HSE and chalcogen abundances in type-A harzburgites. Light grey shaded field encompasses the area of modern MOR harzburgites (Snow and Schmidt, 1998;Luguet et al., 2001, 2003;Harvey et al., 2006;Marchesi et al., 2013); b) Primitive mantle normalised diagram showing HSE and chalcogen patterns of type-B harzburgites. Normalising values are afterBecker et al. (2006),Fischer-Gödde et al. (2011)andWang and Becker (2013).
(Secchiari et al., 2016). Notably, PGE contents and ratios in the lherzolites are similar to other unaltered and serpentinised lherzolites from the modern oceans and ophiolitic complexes (seeFigs. 2 and 3 e.g.;Snow et al., 2000;Luguet et al., 2001, 2004;Pearson et al., 2004; Alard et al., 2005;Becker et al., 2006;Fischer-Gödde et al., 2011; Becker and Dale, 2016). This observation supports the hypothesis that PGE abundances are comparable in fresh and variably serpentinised ul-tramafic rocks (Becker et al., 2006;Becker and Dale, 2016; Fischer-Gödde et al., 2011;Liu et al., 2009;Marchesi et al., 2013), implying that serpentinisation results in minor changes in PGE ratios.
By contrast, the possible influence of serpentinisation on Au and Re is more difficult to evaluate, as no study has systematically investigated its effect on the behaviour of the aforementioned elements. In the lherzolites from New Caledonia, Au displays similar normalised concen-trations as Re and, with the exception of a few samples, both elements are depleted relative to Pd, Te and Se. Au abundances tend to be some-what lower than abundances in other lherzolites with similar major el-ement composition (Fig. 3). Although the compositions can be entirely explained by magmatic fractionation processes (see subsequent chap-ters), minor losses of Au due to hydrothermal alteration cannot be ruled out (e.g.Lorand et al., 1999). The lack of correlation with Al2O3
and YbN(not shown) could be a hint that Au abundances may have
been affected by a combination of magmatic processes and serpentinisation (i.e.Fischer-Gödde et al., 2011). Rhenium is slightly de-pleted compared to the PGE for most of the studied lherzolites, but dis-plays higher concentrations than other mantle lherzolites (Fig. 3). In addition, Re contents do not correlate with LOI and Re/Os ratios cover the range generally reported for moderately depleted mantle rocks. The samples with the lowest Re contents display the highest LOI values, suggesting that no significant quantities of Re were added during the in-teraction with seawater during serpentinisation, as also supported by the chondritic to subchondritic187Re/188Os ratios (seeTable 1).
Likewise, the chondritic to slightly suprachondritic187Os/188Os
can-not be ascribed to serpentinisation, as unrealistically high water–rock ratios (~ 103–104) would be required in order to perturb the whole
rock187Os/188Os at the % level or higher (e.g.Becker and Dale, 2016).
The lherzolite data also show mass fractions of Se and Te and Se/Te that are similar to values in unserpentinised lherzolites (e.g.,Wang and Becker, 2013). In contrast, sulphur in most lherzolites from New Caledonia shows much higher concentrations than typical for perido-tites, which is readily explained by contamination with seawater-derived sulphur during serpentinisation.
We thus conclude that the HSE (perhaps with the exception of Au), Se, Te and Re\\Os signature of the lherzolites offer no evidence to sup-port that serpentinisation and associated reactions affected these ele-ments in a noticeable way.
5.1.2. Partial melting and chemical disequilibrium of the HSE in the mantle Spinel and plagioclase lherzolites exhibit comparable HSE contents and patterns, similar to other lherzolites from oceanic or continental settings that underwent low to moderate degrees of melt extraction (Figs. 2 and 3).
Partial melting has often been invoked as a possible cause for HSE and187Os/188Os variations in mantle rocks (e.g.Meisel et al., 2001;
Reisberg and Lorand, 1995). Studies of the behaviour of HSE during mantle melting and their abundances in mantle rocks have supported the hypothesis that HSE concentrations in residual peridotites result from sulphide-silicate partitioning during magmatic processes (i.e Becker and Dale, 2016;Brenan et al., 2016and references therein). Ex-perimental studies have also highlighted that at temperatures relevant for mantle processes sulphide liquid and, in special cases, sulphide solid solutions, coexist in equilibrium with silicate melt, olivine, pyrox-enes and an Al-rich phase (e.g.Brenan et al., 2016;Mungall and Brenan, 2014;Rehkämper et al., 1999). Experimentally determined sulphide melt-silicate melt partition coefficients (Dsulph/sil) for PGE have been
shown to be high and constant (105to 106, e.g.Mungall and Brenan,
2014;Brenan et al., 2016), while Au shows slightly lower Dsulph/sil
(~104). Therefore, up to moderately high degrees of melting, PGE
be-have as compatible elements and their inter-elemental ratios remain similar as long as sulphide is present in the mantle residue (i.e. until 20–25% of partial melting, depending on initial S content of the source, seeFonseca et al., 2011;Mungall and Brenan, 2014). By contrast, Re is much less chalcophile (Dsulph/sil~300–800, e.g.Fonseca et al., 2007)
and is expected to become more quickly depleted in the residual mantle.
Major element compositions of the sulphide phase observed in the spinel lherzolites (Table S1) are consistent with a residual origin after incongruent melting processes (e.g. Bockrath et al., 2004; Ballhaus et al., 2006). The occurrence of homogeneous monosulphides also sug-gests relatively high cooling rate after the melting event.
Overall, the studied lherzolites are characterised by flat to gently sloping PGE patterns, with similar PM-normalised abundances, no PPGE fractionation and nearly constant ratios for IPGE (i.e. Os/Ir, Ru/ Ir). By contrast, Au and Re display the strongest depletion. These
Fig. 7. a) Fe3+/ΣFe ratios in spinel and b) log(f
O2) ΔFMQ vs Cr# in spinel for samples from this study. Error bars in a = relative error expressed as % (seeTable 2). Error for log(fO2) ΔFMQ
includes error related to fO2sensor (0.5 log units) plus uncertainties derived from Fe3+/Fetotal ratio measurements (seeTable 2). Background fields represent literature data for mantle
peridotites: pink = abyssal peridotites (Bryndzia and Wood, 1990); grey = arc xenoliths (Brandon and Draper, 1996;Parkinson et al., 2003;Wood and Virgo, 1989); light green and light yellow = Toroshima and Conical seamount respectively (Parkinson and Pearce, 1998). Modified afterBirner et al. (2017). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
features imply that HSE, with the exception of Au and Re, exhibit a sim-ilar compatible behaviour during mantle melting, as expected for low to moderate melting degrees in presence of residual sulphide melt. This observation is consistent with the previous estimates obtained through geochemical modelling (Secchiari et al., 2016) and with the occurrence of a residual subsolidus sulphide assemblage in spinel lherzolites.
Although the PGE patterns are nearly flat, with only slight depletion of Pd in a few samples, the depletion of Au and Re, the range of chon-dritic to slightly suprachonchon-dritic γ187Os
iand the higher mass fractions
of Se and Te compared to Re and the other HSE suggest a multi-stage history of the lherzolites. Notably, γ187Os
ido not correlate with mass
fractions of incompatible HSE such as Re, Re/Os nor with fertility indica-tors (Fig. 4), as was observed in some other suites of lherzolites (Becker and Dale, 2016).
The Os isotopic signature may be a pre-existing feature of the mantle source, i.e. already present before the recent melt extraction event (Secchiari et al., 2016), as supported by the dispersed Os isotopes-fertility indicators trends. In addition, the remarkable absence of mag-matic Cu-Fe-rich sulphides (e.g. seeLorand et al., 2003) argue against a recent, post-melting sulphide addition. We thus speculate that the bulk HSE, Se, Te and Os isotope compositions of the lherzolites are the result of partial melting event which affected a mantle source previ-ously characterised by an heterogeneous sulphide population including both residual and magmatic sulfides precipitated along grain bound-aries by infiltrating melts (Burton et al., 1999;Lorand et al., 1999; Alard et al., 2000;Alard et al., 2002)
5.2. Sulphur, Se and Te behaviour in the New Caledonia lherzolites Sulphur mass fractions are variable in the lherzolites from New Cal-edonia and typically much higher compared to estimates of the de-pleted MORB mantle source (DMM ~ 150–200 ppm,Mathez, 1976; Salters and Stracke, 2004) and unaltered lherzolites (e.g.,Wang and Becker, 2013). In addition, sulphur does not correlate with fertility indi-cators (i.e. Al2O3), as commonly observed in unserpentinised mantle
tectonites (e.g.,Lorand and Alard, 2010;Wang and Becker, 2013). The high S concentrations and the lack of correlation with melting indicators suggest that S experienced a late addition during the evolution of these rocks. Sulphur budget of mantle peridotites can be strongly affected by seawater-rock interaction, because of the high sulphate content of sea-water, leading to precipitation of hydrothermal sulphides and sulphate (Alt and Shanks, 1998). We note that the major element chemistry indi-cates a residual origin for the sulphide phases of the lherzolites (see par-agraph 4.1 and 5.1.2). Hydrothermal sulphides or sulphates were not identified during microprobe analyses.
Recent geochemical works have demonstrated the role of serpentine as a sink of S under various oxidation states (S2−, S−, S0and S6+,Debret
et al., 2017). These studies have shown that S concentrations can be anomalously high in serpentinised peridotites (up to 1 wt%, seeAlt and Shanks, 2003), as S can be accommodated in serpentine minerals, accounting from 60 to 100% of the sulphur budget of the peridotite (Debret et al., 2017). In situ XANES analyses have also revealed that S can be hosted in nano-phases associated with serpentine or trapped ei-ther via Si substitutions in the tetrahedra, or as a sulphate ion in the net-work of the tetrahedral sheet of serpentine minerals (Debret et al., 2017).
The addition of S during serpentinisation is also reflected in the high suprachondritic S/Se ratios (up to 16,500) and the excellent correlation observed between S concentrations and S/Se ratios (seeFig. 5c). Despite the strong S enrichments, Se and Te display ‘normal’ concentrations and Se/Te ratios are in the range of other lherzolites (seeWang and Becker, 2013). These data confirm that Se\\Te contents and ratios were not sig-nificantly impacted by serpentinisation, as previously observed for other peridotites that experienced low to moderate serpentinisation de-grees (e.g.Marchesi et al., 2013;Wang and Becker, 2013).
Moreover, Se\\Te show a good correlation between each other (Fig. 5a), implying that they are controlled by the same mineral phases. In mantle peridotites, Se can replace S as a chalcogen anion within the crystalline structure of sulphides (e.g.Bulanova et al., 1996;Hattori et al., 2002;Helmy et al., 2010) or can form Se-rich micro phases, while Tellurium, owing to its semi metal nature, tends to partition be-tween sulphides and late exsolved micrometric tellurides (Pt, Pd, Te, As, Bi phases). The latter are thought to crystallise at low temperatures during cooling, once sulphide melt becomes saturated with respect to Te (Lorand et al., 2008;Lorand and Alard, 2010;Luguet et al., 2004). In the lherzolites, Se and Te do not correlate with melting indicators, but Te displays a rough positive correlation with Pd (Fig. 5b), which sug-gests that the sulphide melt-bulk silicate partition coefficient of Te should be between Se and Pd (e.g.,Figs. 3, 4, 5).
5.3. Type-A harzburgites: highly siderophile element systematics of a resid-ual sub-arc mantle section
The distinct HSE patterns and Os isotopic signature testify that the New Caledonia harzburgites recorded a different evolution compared to the northern lherzolites.
Three of the four samples belonging to the sub-group A (TI1, TI2, PO4) show similar HSE patterns and chalcophile elements depletion, suggesting that the same processes contributed to the HSE and chalcophile element budget of these rocks. The low chalcophile element concentrations, close to or below the detection limit, coupled with low Pd/Ir ratios and subchondritic187Os/188Os
i, point out that type-A
harzburgites are residues after high degree of melt extraction, where sulphides melts must have been nearly completely dissolved in the coexisting silicate melt.
Experimental studies have predicted a compatible behaviour for all the PGE during mantle melting as long as the sulphide phase is retained in the peridotite (Mungall and Brenan, 2014). Depending on the initial S content of the mantle rocks, ~17–20% of melting is required for sulphide exhaustion (Fonseca et al., 2011;Mungall and Brenan, 2014). As melting proceeds, sulphides are progressively dissolved into the melt and the PGE concentrate in the residual sulphide melt (Mungall and Brenan, 2014). Assuming that chemical equilibrium is reached, subsequent melt increments should lead to a slightly increase of whole rock PGE contents, leaving elemental ratios almost constant. At the point when sulphide is completely removed, IPGE and Pt are accommodated in me-tallic alloys, while Re, Au and Pd mass fraction should become extremely low, as these elements are not hosted in any residual mantle phase (Mungall and Brenan, 2014). Hence, for high melting degrees, HSE abundances in the residue should reflect mineral-melt partitioning and the P-T and fO2-dependent solubility of Pt and IPGE alloys in silicate
melt (Brenan et al., 2016;Fonseca et al., 2011, 2012;Mungall and Brenan, 2014).
The IPGE-PPGE fractionation and the resolvable fractionations be-tween specific PGE displayed by the type-A harzburgites bear witness of high melt extraction degrees, which resulted in the formation of a S-free mantle residue. The fractionated Os-Ir-Ru-Rh segments in the HSE patterns and the positive Pt anomalies are likely carried by tiny re-sidual sulphides (i.e. laurite) and metallic alloys (Os\\Ir and Pt\\Ir, see Lorand et al., 1999;Luguet et al., 2001, 2007). These latter are thought to precipitate from sulphide melt shortly before the complete exhaus-tion of sulphide (Mungall and Brenan, 2014) or immediately after sul-phide consumption, due to fS2 lowering and diminished
metal-sulphide complexation in the silicate melt (Fonseca et al., 2012). The variable but broadly systematic IPGE inter-elemental fractionation (high Os/Ir, Ru/Ir and Ru/Rh) and the occurrence of positive Pt anoma-lies possibly suggest the presence of different residual Ir\\Pt alloy pro-portions and preferred Os\\Ru retention compared to Ir in the residual PGE alloys (e.g.Brenan and Andrews, 2001;Fonseca et al., 2012).
The HSE fractionations observed for type-A harzburgites are differ-ent from HSE patterns of modern harzburgites in MOR environmdiffer-ents (Fig. 6a), as the latter are characterised by flat or weakly-fractionated Os-Ir-Ru triplet, rarely displaying positive Pt spikes (Snow and Schmidt, 1998;Luguet et al., 2001, 2003). By contrast, HSE elemental fraction-ations of type-A harzburgites resemble those observed for some arc xe-noliths (Liu et al., 2015;Saha et al., 2005;Scott et al., 2019) or ophiolitic peridotites bearing a supra-subduction zone affinity (seeBüchl et al., 2004,2002;O'Driscoll et al., 2012). Notably, similar HSE fractionations have also been reported for mantle xenoliths from the Chatam Islands (New Zealand, seePearson et al., 2004).
Accordingly, trace element geochemical modelling has shown that the extreme depletion in trace element contents displayed by the New Caledonia harzburgites was achieved through a polyphase evolution, including a first melting event in a mid-ocean ridge setting, followed by fluid-assisted melting reaching clinopyroxene exhaustion in a sub-duction system (seeSecchiari et al., 2020). Such high melting degrees are permissible in supra-subduction zone environments, where hy-drous conditions at relatively low pressures can produce melt frac-tions substantially exceeding 20% without invoking extremely high temperature (e.g. seeUlmer, 2001). In addition, LREE and FME (Sr, Ba, Pb) enrichments coupled with variable Pb isotope compositions of the type-A harzburgites may be explained by syn- and post-melting interactions with different subduction-related components (i.e. aqueous fluids and melts originated in the forearc setting, see Secchiari et al., 2019, 2020).
We thus conclude that the HSE and chalcophile element signature displayed by TI1, TI2 and PO4 predominantly reflect high degrees of melt extraction in a supra-subduction zone environment. The positive Pt spikes suggest that Pt-rich alloys were stable in the mantle residue and were only in part dissolved in the melt during melt extraction.
The enrichments of Au are modest (0.2–1.3 ng/g Au) and may be re-lated to fluid overprint, either from slab-derived fluids (Kepezhinskas et al., 2002;McInnes et al., 1999) or from low-T alteration, e.g. during obduction (e.g.Snow et al., 2000).
The harzburgite YA1 shows higher S and Se concentrations (Fig. 5a), which, considering the significant LOI value of 6.83 wt%, could be related to serpentinisation and precipitation of secondary sulphides. The strongly fractionated HSE pattern and the low concentrations of the in-compatible HSE (i.e. Pd, Re) indicate that the HSE budget of YA1 is also governed by melting, as for TI1, TI2 and PO4 harzburgites. The low Os, Ir and Pt concentrations, are much closer to the values reported for type-B harzburgites (seeTable 1andFig. 5b). Sample YA1 can thus be seen as transitional between type-A and type-B sub-group.
5.4. Origin of type-B harzburgites– strong depletion followed by subduction zone metasomatism?
Type-B harzburgites mostly occur in the central massifs, however one sample (PO3) has also been identified in the eastern zone, close to the area where one type-A harzburgite (PO4) was sampled. Type-B harzburgites display remarkably different HSE and Re/Os behaviour compared to type-A sub-group (Figs. 2 and 6).
Type-B harzburgites show low abundances of incompatible chalcophile elements, i.e. Pd, S, Se and Re, with values in the range of type-A harzburgites. In principle such low concentrations could be rec-onciled with high melting degrees and sulphide exhaustion in the man-tle source. The strong depletion of Os, Ir and Pt, (Fig. 6b) coupled with slightly subchondritic to suprachondritic187Re/188Os (0.23 to 37) and 187Os/188Os (0.1239 to 0.302) are remarkable and do not occur in
resid-ual harzburgites from convecting mantle, as represented by abyssal pe-ridotites. By contrast, type-B harzburgites share some similarities with mantle xenoliths from arc settings, such as low Os contents associated with chondritic to suprachondritic187Os/188Os (Brandon et al., 1996,
1999;Saha et al., 2005;Widom, 2011). These features have been as-cribed to interactions with subduction zone fluids (i.e. fluids from
subducted altered oceanic crust and/or its sedimentary cover), which may have induced sulphide breakdown, locally overprinting the Os iso-topic signature of the mantle (Widom et al., 2003). Such qualitative ob-servations have also been supported by experimental works, that highlighted the critical influence of oxygen fugacity (fO2) on sulphide
and alloy stability (e.g.Andrews and Brenan, 2002;Fonseca et al., 2011, 2012;Mungall and Brenan, 2014). In addition, the strongly frac-tionated, IPGE-depleted, HSE patterns of the type-B harzburgites closely resemble those observed in some refractory harzburgites and replacive dunites from ophiolitic complexes that underwent interaction with S-undersaturated melts (Büchl et al., 2002;Lorand et al., 2004).
As metasomatism by subduction fluids and hydrous melts has been proposed for the New Caledonia harzburgites based on isotopic and in-compatible element studies, we have determined oxygen fugacities on a set of five harzburgites, in order to test if the HSE signature of the two sub-groups could reflect different oxygen fugacity conditions.
As a whole, the harzburgites in our dataset bear witness of similar oxygen fugacity conditions, showing no significant difference among sub-types A and B. Both groups register oxygen fugacity values close to or only slightly higher than the FMQ buffer, displaying a nearly hori-zontal trend in the Cr# vs log (fO2) ΔFMQ variation diagram (Fig. 7b).
Partial melting of a spinel peridotite source is expected to induce progressive Fe3+/ΣFe lowering in the residual mantle, due to the
prefer-ential removal of Fe3+in Al-rich spinel and clinopyroxene during melt
extraction (Woodland et al., 2006). Thus, as melting proceeds and the aforementioned phases are removed from the mantle assemblage, melt extraction in un-buffered conditions should result in fO2lowering
in the residual mantle, generating negative Cr# vs. log (fO2) ΔFMQ
trends (e.g.Brandon and Draper, 1996). However, these correlations are rarely preserved in mantle peridotites, as subsequent geochemical modifications, e.g. metasomatism, tend to decouple fO2from melting
de-pletion indexes. In particular, in the sub-arc region, interaction of the re-fractory peridotite with subduction components is expected to shift the oxygen fugacity state towards higher values, due to the high oxidation capacity of slab-derived fluids and melts (e.g.Parkinson and Arculus, 1999;Brandon and Draper, 1996;Parkinson et al., 2003).
The subhorizontal trends observed inFig. 7a-b, as well as the slightly oxidised fO2values, thus bear evidence that oxygen fugacity was
modi-fied during melting and/or post-melting evolution of the New Caledonia harzburgites. The measured fO2are similar to those observed for the less
oxidised peridotites from the Izu-Bonin region, plotting at the lower end of the arc xenoliths domain (seeFig. 7b; Wood and Virgo, 1989; Brandon and Draper, 1996;Parkinson et al., 2003). Such oxygen fugacity values most likely reflect limited interaction with slab-derived fluids and melts, in agreement with the geodynamic scenario proposed for the New Caledonia archipelago and the previous geochemical models (seeSecchiari et al., 2020). In fact, Eocene subduction is believed to have started close to or in correspondence of an active oceanic spread-ing center, where hot and young (~ 6–9 My old,Cluzel et al., 2016) lith-osphere was forced to subduct. In such a context, fluid fluxes from the downgoing slab must have been limited, due to the young age of the subducted material and the intra-oceanic nature of the subduction (Cluzel et al., 2016). In addition, post melting metasomatism involved small fractions (0.5–1%) of depleted (boninitic) melts, which may have not been able to shift mantle oxidation state towards more oxidised conditions. In fact, while large oxygen fugacity variations in mantle peridotites are permissable for low Fe3+contents in spinel,
more and more Fe3+has to be added to cause even a small increase in
fO2when fO2approaches FMQ buffer (Woodland et al., 2006).
The similar fO2values registered by the two harzburgite sub-types
attest that HSE behaviour was not critically influenced by different oxy-gen fugacity conditions. This observation supports the hypothesis that the two types of patterns are not directly linked to a specific process (i.e. higher melt extraction or enrichment degree or interaction with an oxidised component). Rather, the occurrence of two distinct patterns in the harzburgites may reflect an effect of source-control on the HSE
behaviour during the recent evolution, indicating the presence of geo-chemically heterogeneous mantle domains below the New Caledonia archipelago before Eocene subduction.
5.5. Type-B harzburgites: a broader perspective
Despite being similar in terms of chemistry or mineralogy, type-A and B harzburgites possess distinct HSE signatures. In addition, the HSE signatures of type-B harzbugites have not yet been identified in other mantle tectonites, either from modern oceanic lithosphere or ophiolitic complex.
However, similar compositions have been recently reported for some moderately depleted to highly refractory peridotites and mantle xenoliths from New Zealand (Scott et al., 2019). The New Zealand man-tle is composed of isotopically heterogeneous manman-tle fragments with evolutionary histories extending over 2.75 Ga (Os model ages = 0.1–2.75 Ga, with a broad peak at 1.2 Ga), and PGE systematics decoupled from major element compositions (seeLiu et al., 2015; Scott et al., 2019). These features have been explained by accretion of Zealandia lithospheric mantle from amalgamation of genetically
unrelated convecting mantle fragments which were swept together be-neath the Gondwana subduction margin, variably re-melted and later-ally accreted (Scott et al., 2019).
Among the New Zealand peridotite suites, mantle xenoliths from Lake Moana and Chatam Island show HSE patterns that are similar to our dataset (Fig. 8). Lake Moana Cretaceous xenoliths include cpx-free harzburgites that experienced up to 30% melting, while Eocene-aged Chatam Island harzburgites exhibit a less refractory na-ture, as attested by the presence of primary clinopyroxene (up to 1.8% modal,Scott et al., 2016). Such depletion degrees were achieved either by plume melting or hydrous melting in an arc setting for the Lake Moana xenoliths, whereas Chatam Island samples are thought to represent fragments of fore-arc lithophere (Scott et al., 2016, 2019).
Overall, HSE diagrams highlight that New Zealand mantle xeno-liths reproduce with good approximation both patterns observed in our harzburgites, namely type-A and type-B (Fig. 8). IPGE patterns are broadly sub-parallel, with the New Caledonia harzburgites falling within or at the lower range of values displayed by the New Zealand samples, whereas Pt, Pd and Re exhibit greater variability.
Fig. 8. Primitive mantle normalised HSE abundances of a) type-A and b) type-B harzburgites compared to the HSE composition displayed by New Zealand mantle xenoliths (Scott et al., 2019). See text for further detail.
